The continuous research in the technology of video acquisition devices increases the number of applications with best performance and less cost. For object recognition, navigation and surveillance systems, object detection and tracking are the indispensable steps. Object detection means segmentation of images between foreground and background objects. Object tracking establish the correspondence between the objects in successive frames of video sequence. In this paper, we have proposed algorithms consists of two stages i.e. object detection using Gaussian Mixture Model (GMM) and multiple moving objects tracking using Kalman filter. While tracking the moving object, problems occur during occlusion of persons with each other. However, it can be effectively deal with various video sequences such as indoor, outdoor and cluttered scenes. The experimental results shows that proposed algorithm achieve accurate, robust and efficient results for detection as well as for tracking the foreground objects from complex and dynamics scenes.
General TermsSegmentation, Gaussian Mixture Model, Occlusion, Kalman Filter.
In this paper, research carried out to test the wavelet and cooccurrence matrix based features for rotation invariant texture image retrieval using fuzzy logic classifier. Energy and Standard Deviation features of DWT coefficients up to fifth level of decomposition and eight features are extracted from cooccurrence matrix of whole image and each sub-band of first level DWT decomposition. The texture image is rotated in six different angle. Each rotated texture image sampled to the 128x128, and 256x256 size. The suitability of features are tested using a fuzzy logic classifier. The performance is measured in terms of Success Rate. Success rate is calculated for each rotated texture samples and each of the feature sets. The minimum number of features required for maximum average success rate is obtained. The research shows that for samples of 256x256 size, excellent success rate is achieved for all rotation angle with Wavelet Statistical Features (WSF) as well as Wavelet Cooccurrence Features (WCF). Also energy features perform better than standard deviation features for every rotation angle considered. Also worst case analysis shows that energy features never fail to classify for any of the texture image and more consistent than other features, during the experiment. 8 cooccurrence feature set performs better than 5 co-occurrence feature set. For both the types of features performance degrades in case of 128x128 sample size.
Multifunction parallel image processing systems use standard buses to do inter core communication. Faster and scalable approaches are needed to improve the throughput of the system, but for data heavy applications like Image Processing (IP) algorithms there is a need for constant data transfer between different functional blocks on chip. The solution would either be hardwired buses or controlled communication. Networks-On-Chip (NoC) present a systematic solution, and can succeed a hardwired bus solution in a scalable form. This paper presents a multifunction image processing system prototyped on a single reconfigurable platform. The different IP cores have been implemented keeping in mind on-the-fly processing times and frame rates. The different modules are interconnected using a Torus architecture NoC with an information heavy packet structure and capable of addressing multiple nodes simultaneously. The implementation was done using a low cost Spartan 6 FPGA. Frame rates for standard sizes and chip utilization has been reported.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.